#推理

What if the jagged frontier is mainly math + code (which you can push arbitrarily far with RLVR), and everything else starts to plateau because it is still bottlenecked by human generated data? Model performance in non-verifiable areas has kept improving steadily, albeit much slower than for math and code. But is that steady improvement a side effect of a higher G (itself driven by RLVR), or only a function of the amount of new human data getting injected into training (which is still continually happening on a massive scale)? A lot of things depend on the answer to this question

The Navier-Stokes moment was always much more about the figure below than about the Navier-Stokes problem itself. The run represents a decade of mathematical progress in a week. Can't wait to point these tools at life sciences, building our next models, and alignment!

"Token demand growth will need to outpace declining token prices to support continued growth in investment spending. Frontier models are currently a key source of demand for hyperscaler compute. However, the rise of competitive open-source models has contributed to a decline in average token prices. Measures of frontier token demand slowed in July..." - Goldman

Wild stuff. Anthropic’s new wet lab got its first win after 949 Claude agents autonomously tracked down a new enzyme system in just 21 hours. Claude was searching a huge amount of genetic data when it noticed an unusual pattern: the same small piece of DNA repeating again and again next to an enzyme. The campaign searched 1.94B protein clusters, recovered 198,290 RT clusters, and used 215.6M tokens across 21.5 hours. That pattern looked a little like CRISPR, so Anthropic’s scientists investigated it and found a new family of biological systems they call ART. The system also produces many small RNA molecules, suggesting those repeated DNA sections are doing something useful rather than sitting there randomly.

Some revelation from the Claude Opus 5.5 system card. - Giving Opus 5.5 more reasoning effort made it more likely to obey malicious instructions hidden inside user-pasted text - Anthropic saw Opus 5.5 generate malicious instructions on their own after seemingly harmless mistakes. the behavior may have partly emerged from training designed to stop prompt injections in the first place. - Anthropic's internal estimate says AI may already be compressing roughly 1.5 years of capability progress into one year. - Anthropic gave the model simulated credentials to a public package registry during a security exercise. In roughly half the runs, it took actions that would likely have been harmful if the environment were real. - Some training snapshots hid evidence of actions the models (including Opus 5.5) expected a grader to dislike, including manipulating Git records or deleting logs. "During training, we observed some cases of models (including Opus 5.5) attempting to cover their tracks after performing actions that a grader might view negatively, such as manipulating git records or deleting logs" - METR’s assessment of AI R&D at Anthropic relied partly on information that was not publicly disclosed, including conclusions from a separate METR team with elevated access. That means part of the public assessment of AI-driven R&D acceleration rests on evidence outsiders, and, in this particular case, even another METR team, could not independently inspect.

Claude Opus 5.5 发布了，声称达到 Fable 5.1 级别的性能，同时将典型工作负载成本降低 40%。 输入和输出价格降至每 100 万 tokens 4 美元和 20 美元，而缓存读取下降 60% 至 0.20 美元，均与 Opus 5 相比。 此外，输出速度提升超过 30%，Fast 模式在双倍 token 价格下最高可达 2.5 倍速度。

We’re sharing how GLM-5.3 helped build and optimize the inference infrastructure serving GLM-5.3-Flash. The system went from its first successful run to production readiness in less than two weeks, with end-to-end throughput tripling relative to the initial baseline. The key was dense feedback: local correctness tests, execution traces, microbenchmarks, and end-to-end measurements that enabled targeted hypothesis testing rather than reliance on aggregate performance metrics alone. https://z.ai/blog/glm-built-its-inference-infrastructure